agent-browser

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The prompt contains explicit examples that embed plaintext passwords/tokens in commands and form fills (e.g., agent-browser fill @e2 "password123", echo "pass" | agent-browser auth save ...), so an agent automating logins could be required to output secret values verbatim, creating exfiltration risk.

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). The list includes an explicit malicious domain (https://malicious.com) plus several ambiguous/unverified hosts (site-a/site-b and a third‑party install doc for lightpanda) even though some entries are benign pages (example.com, GitHub login, localhost), so the presence of a known-malicious URL and multiple non-official/unvetted domains makes this set a high-risk source for downloads/executables.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). The skill instructs the agent to navigate and scrape arbitrary public websites (e.g., "agent-browser open " and snapshot/get text in SKILL.md and templates like templates/capture-workflow.sh) and to parse/act on that page-derived content (refs, get text, snapshot) which clearly exposes it to untrusted third-party content that can influence subsequent actions.